Transcript The physical symbol system hypothesis

Chapter 6:
The physical symbol system
hypothesis
Mental architectures approach
Starts off from the basic assumption that cognition is a form of
information-processing
Assumption governs all levels of organization (from neurons
upwards) and almost all explanatory models/hypothesis within the
individual cognitive sciences
Aims to integrate the cognitive sciences at different levels of
explanation/organization through
(1) a model of information-processing
(2) a model of functional organization
Mental architecture
A mental architecture is a model of how the mind is
organized and how it works to process information
1) In what format does a cognitive system carry
information?
2) How does that system transform and process
information?
3) How is the mind as a whole organized into informationprocessing sub-systems?
Two models of informationprocessing
The physical symbol system hypothesis
• e.g. Turing machine model of informationprocessing
• associated with classical, symbolic AI
Connectionism/artificial neural networks
• neurally-inspired models of informationprocessing
• used to model cognitive/perceptual abilities that have
posed problems for classical AI
1975 Turing Award
• Given by Association of Computing Machinery to Allen
Newell and Herbert Simon – pioneers of AI
• Logic Theory Machine (1957)
• General Problem Solver (1956)
• Newell and Simon used their Turing lecture to deliver a
manifesto about the basic principles for studying intelligent
information-processing
Laws of qualitative structure
Basic principles governing individual sciences
Biology: The cell is the basic building block of
organisms
Geology: Geological activity results from the
movement of a small number of huge plates
AI/Cognitive Science: The physical symbol
system hypothesis
The physical symbol system
hypothesis
A physical symbol system has the necessary and sufficient
means for intelligent action
Necessity: Anything capable of intelligent action
is a physical symbol system
Sufficiency: Any (sufficiently sophisticated) PSS
is capable of intelligent action
Four basic ideas
(1) Symbols are physical patterns
(2) Symbols can be combined to form complex symbol
structures
(3) The system contains processes for manipulating complex
symbol structures
(4) The processes for representing complex symbol structures
can themselves by symbolically represented within the
system
Thinking and the PSSS
• The essence of intelligent thinking is the ability to solve
problems
• Intelligence is the ability to work out, when confrionted
with a range of options, which of those options best
matches certain requirements and constraints
• Problem-solving is relative to a problem-space
Specifying a problem in AI
• Basic components of a representation
• description of given situation
• operators for changing the situation
• a goal situation
• tests to determine whether the goal has been reached
• Problem space = branching tree of achievable situations
defined by potential application of operators to initial situation
[e.g. chess]
Problem-solving
Problem-spaces are generally too large to be
searched exhaustively (brute force algorithms)
Search must be selective  heuristic search rules
• effectively close off branches of the tree
• e.g. in chess: “ignore branches that start with
a piece being lost without compensation”
Combinatorial explosion!
• With n connected cities there
are (n – 1)! possible paths
through the search space
• This can be reduced to 2n
• But it would take a computer
processing 1,000,000
possibilities per second over
30 years to solve a 50 city TP
problem by brute force search
Heuristic search hypothesis
Problems are solved by generating and modifying symbol
structures until a solution structure is reached
• GPS starts with symbolic descriptions of the
start state and the goal state
• aims to find a sequence of admissible
transformations that will transform the start state
into the goal state
Heuristic search and algorithms
• The PSSH is a reductive characterization of
intelligence
• It is only illuminating if physical symbol
systems are not themselves intelligent
• This means that the physical symbol
systems must function algorithmically
Illustration: Missionary and cannibals
Symbolic representation of state as mcb
m
= number of missionaries on starting bank
c
= number of cannibals on starting bank
b
= number of boats on starting bank
Start state
Goal state
=
=
331
000
Permissible transformations?
Permissible transformations
micibk
either
or
or
or

mi+1ci+1b1-k
where
difference between mi and mi+1 = 2 and
difference between ci and ci+1 = 0
difference between mi and mi+1 = 1 and
difference between ci and ci+1 = 1
difference between mi and mi+1 = 0 and
difference between ci and ci+1 = 2
either the difference between mi and mi+1
= 1 or the difference between ci and ci+1 = 1
Impermissible states
A branch ends if it reaches a state mcb where
c>m
[more cannibals than missionaries
on R bank]
(3 – c) > (3 – m)
[more cannibals on L bank]
mcb has already appeared earlier in the tree
The overall lie of the land
• The language of thought
hypothesis is a specific
proposal for developing the
PSSH
• The example of WHISPER
shows that symbol structures
can be pictorial
• The contrast class for the PSSH
is the class of neural network
(connectionist) models